用神经网络和高光谱植被指数估算小麦生物量  

ANN-based wheat biomass estimation using canopy hyperspectral vegetation indices

作　　者：王大成[1,2] 王纪华 靳宁[1] 王芊[1] 李存军[1] 黄敬峰[2] 王渊[2] 黄芳[2] 

机构地区：[1]国家农业信息化工程技术研究中心,北京100097 [2]浙江大学农业遥感与信息技术应用研究所,杭州310029

出　　处：《农业工程学报》2008年第S2期196-201,共6页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家自然科学基金(40701120);国家高技术研究计划(2006AA12Z138);北京市优秀人才项目(20071D0200500046)

摘　　要：准确的估算作物生物量,能够为国家和地方政府的粮食经济宏观调控决策提供科学依据。利用高光谱植被指数,系统的比较了人工神经网络方法和传统回归模型估算的小麦生物量。结果表明,基于BP神经网络的方法相对于对数回归模型,显著地提高了小麦的生物量诊断的准确性,均方根误差(RMSE)相对减小,决定系数(R2)和T值相对增大,特别是对于比值植被指数(RVI),T值提高的幅度比较大,达99.8%。说明人工神经网络对作物小麦的生物量高光谱遥感诊断是一种实时高效的方法。The accurate estimation of crop biomass gives a scientific guidance for national and local governments to determine the macro-control policy of grain economy.Using hyperspectral Vegetation Indices(VIs),a systematic comparison was conducted between methods of artificial neural network and traditional regression model to estimate crop biomass.The results show that the BP-based artificial neural network method,compared with the logarithm regression model,remarkably enhances the accuracy of wheat biomass estimation,reduces Root Mean Square Error (RMSE)and increases determination coefficient and T-test values relatively.T values are raised sharply to 99.8%, especially for RVI.It is concluded that the artificial neural network is a real-time and efficient way for hyperspectral wheat biomass.

关 键 词：小麦 高光谱 植被指数 人工神经网络 对数回归 

分 类 号：S512.1[农业科学—作物学]